EGPGV18: Eurographics Symposium on Parallel Graphics and Visualization

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2632395

Browse

Now showing 1 - 11 of 11

Hardware-Accelerated Multi-Tile Streaming for Realtime Remote Visualization
(The Eurographics Association, 2018) Biedert, Tim; Messmer, Peter; Fogal, Thomas; Garth, Christoph; Hank Childs and Fernando Cucchietti
The physical separation of compute resource and end user is one of the core challenges in HPC visualization. While GPUs are routinely used in remote rendering, a heretofore unexplored aspect is these GPUs' special purpose video en-/decoding hardware that can be used to solve the large-scale remoting challenge. The high performance and substantial bandwidth savings offered by such hardware enable a novel approach to the problems inherent in remote rendering, with impact on the workflows and visualization scenarios available. Using more tiles than previously thought reasonable, we demonstrate a distributed, lowlatency multi-tile streaming system that can sustain stable 80 Hz when streaming up to 256 synchronized 3840x2160 tiles and can achieve 365 Hz at 3840x2160 for sort-first compositing over the internet.
Dense Texture Flow Visualization using Data-Parallel Primitives
(The Eurographics Association, 2018) Kim, Mark; Klasky, Scott; Pugmire, David; Hank Childs and Fernando Cucchietti
Recent trends in supercomputing towards massively threaded on-node processors to increase performance has also introduced fragmented software support. In response to this changing landscape, new scientific visualization packages have been developed to provide a portable framework to exploit this on-node parallelism with data parallel primitives, while also providing a single interface to multiple hardware backends. This necessitates adapting algorithms to the data parallel primitives paradigm. In numerous cases the algorithm is serial, but other times the technique is tied to hardware and needs to be generalized to broadly disseminate. In this work, we present unsteady flow line integral convolution (UFLIC) using only data parallel primitives. Line integral convolution (LIC) is a fundamental flow visualization technique in scientific visualization. LIC and its texture-based variants, are used in fields such as meteorology and computational fluid dynamics to aid practitioners because of its efficient memory usage, strong, visual flow characteristics, and efficient performance. However, in practice performant implementations are GPU shader-based approaches, which limits deployment and adoption. By utilizing VTK-m, our approach is a performant, memory efficient implementation, with the added benefit of portability, with a single implementation across many architectures.
VisIt-OSPRay: Toward an Exascale Volume Visualization System
(The Eurographics Association, 2018) Wu, Qi; Usher, Will; Petruzza, Steve; Kumar, Sidharth; Wang, Feng; Wald, Ingo; Pascucci, Valerio; Hansen, Charles D.; Hank Childs and Fernando Cucchietti
Large-scale simulations can easily produce data in excess of what can be efficiently visualized using production visualization software, making it challenging for scientists to gain insights from the results of these simulations. This trend is expected to grow with exascale. To meet this challenge, and run on the highly parallel hardware being deployed on HPC system, rendering systems in production visualization software must be redesigned to perform well at these new scales and levels of parallelism. In this work, we present VisIt-OSPRay, a high-performance, scalable, hybrid-parallel rendering system in VisIt, using OSPRay and IceT, coupled with PIDX for scalable I/O. We examine the scalability and memory efficiency of this system and investigate further areas for improvement to prepare VisIt for upcoming exascale workloads.
Interactive Visual Analysis of Multi-dimensional Metamodels
(The Eurographics Association, 2018) Gebhardt, Sascha; Pick, Sebastian; Hentschel, Bernd; Kuhlen, Torsten Wolfgang; Hank Childs and Fernando Cucchietti
In the simulation of manufacturing processes, complex models are used to examine process properties. To save computation time, so-called metamodels serve as surrogates for the original models. Metamodels are inherently difficult to interpret, because they resemble multi-dimensional functions f : Rn -> Rm that map configuration parameters to production criteria. We propose a multi-view visualization application called memoSlice that composes several visualization techniques, specially adapted to the analysis of metamodels. With our application, we enable users to improve their understanding of a metamodel, but also to easily optimize processes. We put special attention on providing a high level of interactivity by realizing specialized parallelization techniques to provide timely feedback on user interactions. In this paper we outline these parallelization techniques and demonstrate their effectivity by means of micro and high level measurements.
Rapid k-d Tree Construction for Sparse Volume Data
(The Eurographics Association, 2018) Zellmann, Stefan; Schulze, Jürgen P.; Lang, Ulrich; Hank Childs and Fernando Cucchietti
While k-d trees are known to be effective for spatial indexing of sparse 3-D volume data, full reconstruction, e.g. due to changes to the alpha transfer function during rendering, is usually a costly operation with this hierarchical data structure. We pick a serial state of the art implementation that is based on summed-volume tables and propose a parallel version of the construction algorithm for multi-core CPUs. Our parallel k-d tree construction algorithm can be used to rapidly perform full hierarchy rebuilds for moderately sized to large volume data sets. We reformulate the original, highly serial construction algorithm by replacing the summed-volume table (SVT) that is used to perform fast occupancy queries with a list of partial summed-volume tables. This gives rise to parallelism at several stages of the algorithm. We show how to achieve high scalability with a carefully crafted parallelization scheme. As a side effect, our construction algorithm also relaxes the tremendous memory overhead imposed by full SVTs. For our scalability study, we have integrated the parallel k-d tree implementation into a ray casting volume rendering pipeline. We present comparisons for various sparse 3-D volumetric data sets where k-d trees are first built interactively and then later used to skip over empty space.
Robust Iterative Find-Next-Hit Ray Traversal
(The Eurographics Association, 2018) Wald, Ingo; Amstutz, Jefferson; Benthin, Carsten; Hank Childs and Fernando Cucchietti
We present two different methods for improving the performance and robustness of ray tracing algorithms that require iterating through multiple successive intersections along the same ray. Our methods do so without ever missing intersections, even in cases of co-planar or numerically close surfaces, and without having to rely on a dedicated all-hit traversal kernel. Furthermore, our evaluation shows that, in many cases, our methods consistently outperform existing all-hit kernels.
Performance-Portable Particle Advection with VTK-m
(The Eurographics Association, 2018) Pugmire, David; Yenpure, Abhishek; Kim, Mark; Kress, James; Maynard, Robert; Childs, Hank; Hentschel, Bernd; Hank Childs and Fernando Cucchietti
Particle advection is the fundamental kernel behind most vector field visualization methods. Yet, the efficient parallel computation of large amounts of particle traces remains challenging. This is exacerbated by the variety of hardware trends in today's HPC arena, including increasing core counts in classical CPUs, many-core designs such as the Intel Xeon Phi, and massively parallel GPUs. The dedicated optimization of a particle advection kernel for each individual target architecture is both time-consuming and error prone. In this paper, we propose a performance-portable algorithm for particle advection. Our algorithm is based on the recently introduced VTK-m system and chiefly relies on its device adapter abstraction. We demonstrate the general portability of our implementation across a wide variety of hardware. Finally, our evaluation shows that our hardware-agnostic algorithm has comparable performance to hardware-specific algorithms.
La VALSE: Scalable Log Visualization for Fault Characterization in Supercomputers
(The Eurographics Association, 2018) Guo, Hanqi; Di, Sheng; Gupta, Rinku; Peterka, Tom; Cappello, Franck; Hank Childs and Fernando Cucchietti
We design and implement La VALSE-a scalable visualization tool to explore tens of millions of records of reliability, availability, and serviceability (RAS) logs-for IBM Blue Gene/Q systems. Our tool is designed to meet various analysis requirements, including tracing causes of failure events and investigating correlations from the redundant and noisy RAS messages. La VALSE consists of multiple linked views to visualize RAS logs; each log message has a time stamp, physical location, network address, and multiple categorical dimensions such as severity and category. The timeline view features the scalable ThemeRiver and arc diagrams that enables interactive exploration of tens of millions of log messages. The spatial view visualizes the occurrences of RAS messages on hundreds of thousands of elements of Mira-compute cards, node boards, midplanes, and racks-with viewdependent level-of-detail rendering. The multidimensional view enables interactive filtering of different categorical dimensions of RAS messages. To achieve interactivity, we develop an efficient and scalable online data cube engine that can query 55 million RAS logs in less than one second. We present several case studies on Mira, a top supercomputer at Argonne National Laboratory. The case studies demonstrate that La VALSE can help users quickly identify the sources of failure events and analyze spatiotemporal correlations of RAS messages in different scales.
Direct Raytracing of Particle-based Fluid Surfaces Using Anisotropic Kernels
(The Eurographics Association, 2018) Biedert, Tim; Sohns, Jan-Tobias; Schröder, Simon; Amstutz, Jefferson; Wald, Ingo; Garth, Christoph; Hank Childs and Fernando Cucchietti
Particle-based simulation models have assumed a significant role in the numerical computation of high-fidelity transient flow and continuum mechanical problems. However, direct visualization of surfaces from particle data without intermediary discrete triangulation remains a challenging task. We demonstrate a novel direct raytracing scheme for free surface intersection based on anisotropic smoothing kernels. Our approach efficiently reduces the number of candidate kernels evaluated to converge to the surface threshold, thereby running in image space rather than object space complexity. We conduct comprehensive benchmarks with respect to data set size, scene complexity, visual fidelity and hardware setup. Our versatile system is suitable for both high quality and interactive desktop rendering, scales reasonably well even with trivial parallelization and renders up to 170 million particles on 32 distributed compute nodes at close to interactive frame rates at 4K resolution with ambient occlusion.
Revisiting the Evaluation of In Situ Lagrangian Analysis
(The Eurographics Association, 2018) Sane, Sudhanshu; Bujack, Roxana; Childs, Hank; Hank Childs and Fernando Cucchietti
In situ usage of Lagrangian techniques has proven to be superior with respect to emerging supercomputing trends than the traditional Eulerian approach for scientific flow analysis. However, previous studies have not informed two key points: (1) the accuracy of the post hoc interpolated trajectory as a whole and (2) the spatiotemporal tradeoffs involved when using Lagrangian analysis. With this short paper, we address these points. We first conduct a more comprehensive evaluation via additional accuracy metrics tailored for evaluating Lagrangian trajectories. Second, we provide an understanding of the configurations where the Lagrangian approach works well by studying spatiotemporal tradeoffs. In addition, our study highlights the effects of error propagation and accumulation when performing Lagrangian interpolation for large numbers of steps. We believe our study is significant for better understanding the use of in situ Lagrangian techniques, as well as serving as a baseline for future Lagrangian research.
EGPGV 2018: Frontmatter
(The Eurographics Association, 2018) Childs, Hank; Cucchietti, Fernando; Hank Childs and Fernando Cucchietti

Browse

Browsing EGPGV18: Eurographics Symposium on Parallel Graphics and Visualization by Issue Date

Results Per Page

Sort Options